Identifying Urban Typologies through remote sensing and clustering

Navigating the diverse definitions of urban and rural areas, along with the varied urban patterns within cities, poses a significant challenge when comparing cities globally. Different regions or countries employ distinct combinations of physical and socioeconomic attributes to define a city, complicating efforts to establish uniform criteria. While organizations like the OECD, the EU, or the US Census Office propose urban definitions, researchers analyzing global urban patterns, especially in the global south, often grapple with the need to amalgamate data from multiple sources. This amalgamation introduces a high level of variability in spatial boundaries, making it challenging to directly apply established criteria from one region to differentiate between predominantly urban and non-urban neighborhoods in another.

Building on our previous experience evaluating and comparing urban areas’ environmental performance, we present an approach that could help researchers in this field to achieve a more accurate differentiation between urban and non-urban areas using a combination of geospatial analysis and traditional clustering tools applied to two use cases. The first one, developed as part of the Data-Driven EnviroLab’s Urban Environment and Social Inclusion project, uses a combination of physical and socioeconomic variables at the neighborhood level and a stepwise hierarchical clustering to identify groups of neighborhoods that are not functionally urban. The second approach uses a similar set of variables but at the city level and k-clustering to identify groups of cities with similar development patterns and context, in order to analyze their attributes within and across groups. The application of this approach has allowed us to both reduce the variability within each city and create meaningful groups of cities/neighborhoods which can be used for more accurate within and across group comparison, and has potential to serve as a viable method to segment large urban databases for creating more refined analysis for multiple uses.